Felix Ocker, Jörg Deigmöller, Julian Eggert, "Exploring Large Language Models as a Source of Common-Sense Knowledge for Robots", International Semantic Web Conference, 2023.

By definition, service robots are supposed to help humans in everyday situations. To behave as expected, the robots require a sound knowledge base, allowing them to infer necessary actions. For situations such as serving a drink in the desired way, common-sense knowledge is required. The challenge with common-sense knowledge is that it is inherently implicit, i.e., it is self-evident for humans but not explicitly documented. Compared to the amount of potentially relevant knowledge, there are only very little explicit sources, created in cumbersome community efforts. With the rise of large language models, a potential source for such knowledge becomes available. This paper explores whether recent large language models can serve as a source of common-sense knowledge for robots and compares their performance to traditional approaches based on databases created in community efforts.